Journal Article

MOAEOSCA: an enhanced multi-objective hybrid artificial ecosystem-based optimization with sine cosine algorithm for feature selection in botnet detection in IoT

Gharehchopogh, Farhad Soleimanian, Hosseini, Fatemeh, Masdari, Mohammad

2023

Overview

The number of Internet of Things (IoT) devices overgrows, and this technology dominates. The importance of IoT security and the growing need to devise intrusion detection systems (IDSs) to detect all types of attacks. The limited sources on the IoT. They have led researchers to explore and provide new and efficient solutions to create Botnet Detection in IoT systems. These systems use data features to detect network traffic status and thus detect malicious behavior. Also, data set features indicate the type of network traffic. Many features in the problem space and network behaviour unpredictability make IDSs the main challenge in establishing security in computer networks. Many unnecessary features have also made feature selection an essential aspect of attack detection systems. This paper developed a multi-objective MOAEOSCA algorithm hybridizing Artificial Ecosystem-based Optimization (AEO) algorithms and the Sine Cosine Algorithm (SCA) for botnet detection in IoT. By accurately identifying the weaknesses of the MOAEOSCA algorithm, it has been tried to cover the weaknesses to a large extent and to reach a robust algorithm. We promoted the proposed algorithm using Bitwise operations, Disruption operator, and Opposition-based learning (OBL) mechanisms. Ten standard datasets in the UCI repository were examined to evaluate the proposed algorithm’s performance in solving the feature selection problem to detect a botnet. Simulation findings indicated that the proposed algorithm had an acceptable accuracy in Botnet Detection in the IoT, outperforming other methods. According to the experiments carried out in this paper, the MOAEOSCA algorithm has shown that nine data sets out of ten data sets in the feature selection problem performed better than other optimization algorithms. But in all seven botnet data sets, performance has shown better than different optimization algorithms.

• Publisher: Springer Nature B.V
• Subject: Algorithms / Communications traffic / Computer networks / Cybersecurity / Datasets / Feature selection / Internet of Things / Malware / Multiple objective analysis / Optimization / Optimization algorithms / Trigonometric functions